subtopic

PCA: variance, components, loadings and reconstruction

Fit PCA; plot cumulative explained variance; read loadings (A); reconstruct digits from 5, 10, 20 components (B). Concepts: Principal components; explained variance; loadings; projection; reconstruction error Project A: PC1 ≈ overall spend, PC2 ≈ category preference (e.g. home décor vs gifts) — check loadings. Project B: ~20 components keep ~90% variance; reconstructed digits stay readable.

Builds on: B19 · Build: See Your Data in 2-D (PCA)